Michał Brzoza-Brzezina, Marcin Kolasa, Mateusz Szetela Is Poland at risk of the zero lower bound?

(1)

Bank i Kredyt 47(3), 2016, 195-226

Is Poland at risk of the zero lower bound?

Michał Brzoza-Brzezina

*

_{, Marcin Kolasa}

#

_{, Mateusz Szetela}

ǂ

Submitted: 20 October 2015. Accepted: 24 May 2016.

Abstract

In early 2015, the policy (open market operations) rate of Narodowy Bank Polski was reduced to an all- -time low of 1.5%. At the same time, prices of consumer goods and services dropped by 1.5% in year-on- -year terms. This raised concerns that Poland might become the next country to hit the zero lower bound (ZLB) constraint on nominal interest rates. The purpose of this paper is to examine the scale of this risk and its possible consequences. According to our results, the odds of the Polish economy hitting the ZLB remain low, despite having risen considerably in 2014−2015. At the same time, the consequences of such a scenario would be substantial as the ZLB would amplify the economy’s responses to adverse demand shocks and make their impact more persistent. The current level of the inflation target (2.5%) protects the Polish economy against the zero lower bound to a signifficant degree. However, its potential reduction would significantly increase the likelihood that this threat materializes.

Keywords: zero lower bound, Polish monetary policy, small open economy JEL: E43, E47, E52

*_{Narodowy Bank Polski and Warsaw School of Economics; e-mail: michal.brzoza-brzezina@nbp.pl.} #_{Narodowy Bank Polski and Warsaw School of Economics; e-mail: marcin.kolasa@nbp.pl.} ǂ _{Warsaw School of Economics; e-mail: mat.szetela@gmail.com.}

(2)

M. Brzoza–Brzezina, M. Kolasa, M. Szetela

196 1. Introduction

Both the literature (discussed in more detail in the following section) and the empirical data from countries which have faced the zero lower bound (ZLB) constraint on nominal interest rates in the recent years show that it may have serious consequences for macroeconomic stability. The ZLB reduces the central bank’s capacity to stabilize macroeconomic conditions, thus boosting inflation and output volatility. In early 2015, the policy rate of Narodowy Bank Polski was reduced to its all-time low of 1.5%. At the same time, the prices of consumer goods and services sank by 1.5% in year-on-year terms. As a result, concerns arose that the Polish economy might be the next in line to hit the ZLB.

In this paper we examine the ZLB phenomenon in the context of the Polish economy. In particular, our aim is to assess the probability of Poland hitting the ZLB, possible consequences of such a scenario, and its relationship with the level of the inflation target. To this end, we use a dynamic stochastic general equilibrium (DSGE) model of a small open economy, estimated with Bayesian methods on data for Poland and the euro area.

The main findings of our study are as follows. First, the probability of Poland hitting the ZLB is relatively low. However, while it was virtually zero until 2013, it has begun to rise sharply more recently. Still, even when deflationary processes were at their most severe at the turn of 2014 and 2015, the probability of the ZLB scenario unfolding in Poland in a 3-year horizon did not exceed 6%. Second, the current inflation target is high enough to provide good insurance against the ZLB as the estimated median time to hit the ZLB, starting from the steady state, is over 100 years. However, a reduction of the target would significantly increase the risk of hitting the zero lower bound. For example, if the target was 1.5%, the economy would reach the ZLB after an average of only 22 years, with a 25% probability of this risk materializing in less than seven years. Third, the impulse response analysis shows that the economy confronted with the zero lower bound reacts more weakly to positive supply shocks and more strongly to negative disturbances to consumption preferences, government expenditure, risk premia or external shocks. Fourth, in line with findings from the previous literature, the effectiveness of fiscal stimulus increases significantly under the ZLB: the government expenditure multiplier may be even twice as high as in normal times. Fifth, the spill- -over of a crisis in the euro area to Poland is, ceteris paribus, smaller if the euro area is stuck at the ZLB.

The rest of the paper is divided into 5 sections. In Section 2, we discuss briefly the ZLB problem. Sections 3 and 4 present the model and its estimation. In Section 5, we discuss the results of our simulations. Section 6 concludes.

2. The zero lower bound

The zero lower bound is said to be binding when the central bank has reduced the policy rate to near zero, whereas the current and anticipated macroeconomic situation call for further cuts. In this section we discuss briefly two key issues related to the ZLB: the ways to avoid it and the consequences of being trapped. Both are closely related to our simulations presented in the next sections. For the sake of brevity we do not discuss unconventional monetary policy instruments that can be used at the ZLB. Interested readers can see, among others: D’Amico and King (2010), Greenwood and Vayanos (2008),

(3)

Is Poland at risk of the zero lower bound?

197

Krishnamurthy and Vissing-Jorgensen (2011), Joyce et al. (2011), Gagnon et al. (2011), Hamilton and Wu (2012), Gambacorta, Hofmann and Peersman (2012), Engen, Laubach and Reifschneider (2015), Chen et al. (2015) or Woodford (2012).

2.1. Avoiding the ZLB

A key question related to the ZLB is, obviously, how to avoid it. The literature has proposed a number of methods to reduce the risk of hitting the ZLB. Below we concentrate on three that are closely related to the conduct of monetary policy: changing the inflation target, modifying the monetary policy strategy and changing the degree of aggressiveness of monetary policy reactions. The vast literature on directly relaxing the constraints generated by the existence of cash is not discussed here. See e.g. Buiter and Panigirtzoglou (2003), Buiter (2009), Mankiw (2009) or Ilgmann and Menner (2011). For more details on this issue, see also Adam and Billi (2006, 2007), Nakov (2008) or Svensson (2003).

The first way to insure against the ZLB is to change the inflation target. As argued by Blanchard, Dell’Ariccia and Mauro (2010), raising it should lead to lower probability of the ZLB. This happens because a higher target means (over a longer horizon, once the adjustment processes have been completed) permanently higher inflation expectations. For a given natural interest rate (expressed in real terms, which we assume not to depend directly on monetary policy), higher inflation expectations raise the nominal equilibrium rate. Hence, a higher inflation target implies higher nominal interest rates and a larger buffer, allowing for deeper cuts in the nominal interest rate when adverse shocks hit. The problem with the above solution is that higher inflation also involves welfare losses for the economy. The literature provides a long list of reasons why inflation is harmful – the shoe-leather cost, the information cost, the cost of non-optimal allocation of resources or the inflation tax. With this in mind, Coibion, Gorodnichenko and Wieland (2012) formally investigated whether central banks should increase inflation targets in the context of the ZLB. The paper weights a potentially large, yet incidental cost of the ZLB against a relatively small but more permanent cost of higher average inflation. Using a New Keynesian model calibrated for the United States, the authors conclude that the optimal rate of inflation is positive, but not exceeding 2%. This suggests that the current inflation targets of central banks already take into account the lower bound on nominal interest rates. It is also worth noting that no central bank has yet decided to increase the inflation target because of the ZLB risk.

The second idea is to change the monetary policy strategy (Billi 2013; Nakov 2008). According to the inflation targeting (IT) strategy, in times of low (high) inflation the central bank promises to raise (reduce) inflation to the target. Price level targeting (PLT) is an interesting alternative to IT. Under PLT, the central bank conducts its monetary policy with a view to keeping price level growth close to an ex ante designated path. As a result, following a period of low inflation, it is necessary to temporarily generate inflation above the trend. This element of the strategy is crucial in the ZLB context. When inflation falls, future inflation expectations rise. This implies a drop in long-term real interest rates, which in turn leads to an expansion in aggregate demand, thus reducing the decline in inflation. As a result, the ex ante probability of hitting the lower bound decreases.

The third way to prevent the ZLB is to increase the aggressiveness of monetary policy. As shown by Adam and Billi (2006, 2007) and Nakov (2008), optimal monetary policy takes a particular form in

(4)

198

the ZLB environment. When the probability of hitting the ZLB rises, the central bank should reduce the interest rates sharply, and in particular more aggressively than it would have done in normal times. In this way the bank generates a temporary (in contrast to permanent, stemming, e.g. from raising the target) rise in inflation expectations, and hence a decrease in real interest rates, which in turn increases the probability of avoiding the ZLB.

2.2. Consequences of hitting the ZLB

The second issue related to the ZLB that we briefly discuss are its possible adverse consequences. Below, we concentrate on two that have received most attention in the literature: the inability to use conventional (i.e. interest rate-based) monetary policy by the central bank and the modified macroeconomic dynamics at the ZLB.

The first problem resulting from the ZLB is that it limits the ability of the central bank to stimulate the economy. The reason is obvious – the central bank, which cannot lower the policy rate below zero, looses its key instrument, which negatively affects macroeconomic stability. Since the outbreak of the financial crisis, when the ZLB started to spread to a number of economies, this issue has been discussed by both theoretical and empirical literature. For example, Ireland (2011) estimated in a New Keynesian framework that, had it not been for the zero bound for interest rates, the 2009 recession in the United States would have been less severe by approximately 1 percentage point. Gust, Lopez-Salido and Smith (2012) estimated a nonlinear DSGE model for the United States and showed that the US GDP was lower by 1% on average over 2009−2011 because the interest rates could not be sufficiently reduced.

The second problem with the ZLB is linked to the change in the economy’s reactions to shocks. When interest rates are at the ZLB, the response to negative (ie. output-reducing) disturbances may become more pronounced, while the response to some positive shocks may be weaker. For instance, Neri and Notarpietro (2014) showed that a positive technological disturbance, which normally leads to an increase in GDP, may result even in a GDP decline at the ZLB. Baürle and Kaufmann (2014) showed that risk premium shocks are significantly amplified at the ZLB. This case is of particular importance for Switzerland, where the inflow of financial capital caused an enormous appreciation pressure, pushing the economy deeper into the trap. Brzoza-Brzezina (2016) showed that amplification of shocks at the ZLB depends to a large extent on the economy’s openness: the consequences of the ZLB are more severe in a closed economy than in a small open economy. Other studies (e.g. Bodenstein, Erceg, Guerrieri 2009; Haberis, Lipinska 2012) showed that international spillovers are stronger when the ZLB binds.

3. The model

To conduct an empirical analysis focused on the Polish economy, we consider a fairly standard DSGE model of a small open economy, founded on the seminal work of Obstfeld and Rogoff (1995) and its later development by Gali and Monacelli (2005). Compared to more recent applications, our model’s structure is slightly richer than that of Justiniano and Preston (2010) and somewhat simplified in comparison with Adolfson et al. (2007) or Christoffel, Coenen and Warne (2008a). The model economy

(5)

Is Poland at risk of the zero lower bound?

199

is populated by households, producers of intermediate and final goods, importers, as well as fiscal and monetary authorities. The rest of the world is treated exogenously and represented by a simple vector autoregressive (VAR) process.

In what follows we present the problems of agents. The Appendix presents a complete list of log- -linearised equations that we use to estimate and simulate the model.

3.1. Households

There is a continuum of households indexed by _{j. A typical household is assumed to maximize the}

objective function:

(

)

= + + 0 1 , , 1 1 , , 0 ₁ ₁ t t j t l t t j t g t c c l E

Σ

(

)

jt t j t j t t j t j _x x x S k k , 1 , , , 1 , = 1 + 1 = = ≡ ≡ – – – – – – – – – – – – – +

( )

l wdj w lt jt, 1 0 1 t, j t t, j W W w w + 1 1 1

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

t t jt jt jt t t t t t t t t tc Px D eB T D i eB i a Rk W l P = 11+ 1 + 11+ *1 1, + + , , + , = + + + + +

(

at 1, t

)

exp

[

(

at 1 t

)

]

t t t t e_y_PB a y , t

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

X , di y yXt, = Xi,t, = = 1 1 0 1 1 0 di y y * t, i, H * t, H 1 t, i t, i t t, i, H zk l y

(

–

)

(

+

)

– – – – t i H t i H t t i H t t h P,, MC y ,, y*,, t i H t H t i H

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

F,,it t t f P eP y t ,i , F t , F t, i , F P P f f + = 1 1 1 , , ,t = Ft Ft F P P

(

)

[

(

)

]

{

i t i y t p t dp t t m mt

}

t max ; ˆ 1 1 yˆ ˆ ˆ ˆ 1 , ˆ= + + + + * t , H t, H t y y y = + * t * t * _P _/_P 1 * * , * , t t t t H t H _e_P y P y = t t t t c x g y~ = + + t, x x t x t xˆ ˆ xˆ + – – – – = 1

( )

0

,

1 N

~

ˆ

x,t t , g , ,lt, t, t, z , t g , t mt,, y , *t *t t y , y~ , t c , t x , t l , t k , t r , t w , t p , kt, t, Ht,, F,t, t, t, mc , t s , t q , t at

(

)

(

)

= = = _h t s t t t h t s t t t d h g g y y M 1 1 1 1 β ε ε ϕ ϕ ξ σ δ δ δ π π π

∫

λ λ σ Ξ Φ Φ φ φ φφ ω η η η η η _η ηη ω φ ε ε –1 ε ε θ Λ Λ α α ε ε – 1 ε ε 0 _t E

Σ

0_t₌₀ E

Σ

δ π δ δ π π π π σ ε π π π ψ ψ ψ η ν ψ ψ π δ π ≡ – σ ε ε ε ε γ ε ρ π π β β π π λ φ

Σ

]

[

( ))

(

]

[

]

( )

(1)

where _c_t denotes consumption and _l_tstands for hours worked. The parameters σ and φ are the inverse

of the elasticities of, respectively, intertemporal substitution and labour supply, whereas _{ξ describes}

the degree of external habit formation in consumption. Households’ preferences are disturbed by

consumption preferences shocks _ε_g,t_{and labour supply shocks ε}_l,t.

Households also own capital _k_t and adjust its supply through investment _x_t. The capital law

of motion is:

(

)

Σ

(

)

( )

l wdj w l_t 1 _j_t, 0 1 t, j t t, j W W w w +1 1 1

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

Σ

0_t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(2)

where S(.) is a function which describes the investment adjustment cost and whose second derivative

in the steady state is S'', while Y_t is a stationary technological investment-specific shock.

Each household _{j provides diversified work services l}_j,_t. The total labour supply is specified by

the following Dixit-Stiglitz aggregator:

(

)

Σ

(

)

( )

l wdj w l_t 1 _j_t, 0 1 t, j t t, j W W w w +1 1 1

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

t t jt jt jt t t t t t t t t tc Px D eB T D i eB i a R k W l P = 11+ 1 + 11+ *1 1, + + , , + , = + + + + +

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

– – – – t i H t i H t t i H t t h P ,, MC y ,, y*,, t i H t H t i H

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

Σ

0_t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(3)

where _λ_w is the wage markup.

In each period each household can reoptimize its wages with probability 1– θ_w. Wages of

the remaining households are indexed according to:

(

)

Σ

(

)

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

Σ

0_t₌₀ E

Σ

]

[

( ))

(

]

[

]

( )

(4)

where _{π– is the inflation target,}

(

)

Σ

(

)

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 H,t F,t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

t max ; ˆ 1 1 yˆ ˆ ˆ ˆ 1 , ˆ= + + + + * t , H t, H t y y y = + * t * t * _P _/_P 1 * * , * , t t t t H t H _e_P y P y = t t t t c x g y~ = + + t, x x t x t xˆ ˆ xˆ + – – – – = ₁

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

λ λ σ Ξ Φ Φ φ φ φφ ω η η η η η η η η ω φ ε ε –1 ε ε θ Λ Λ α α ε ε – 1 ε ε 0 _t E

Σ

0_t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

is inflation, P_t is the price level, while δ_w is a parameter

(6)

200

All assets held by households are one-period. We also assume that households have access to a complete insurance market that allows them to insure against idiosyncratic wage risk. As a result, income of all households is the same despite wage stickiness. Households’ budget constraint can be written as follows:

(

)

Σ

(

)

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at1, t

)

exp

[

(

at1 t

)

]

(

)

+ = 1 1 1 1 1 1 H,t F,t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

x,t t , g , ,lt, t, t, z , t g , t mt,, y , *t t* t y , y~ , t c , t x , t l , t k , t r , t w , t p , kt, t, Ht,, F,t, t, t, mc , t s , t q , t at

(

)

(

)

∫

λ λ σ Ξ Φ Φ φ φ φφ ω η η η η η η η η ω φ ε ε –1 ε ε θ Λ Λ α α ε ε – 1 ε ε 0 _t E

Σ

0 _t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(5)

where _D_t are claims on other domestic households, _B_t are claims on foreign households, _e_t denotes

the exchange rate, _R_t is the rental rate on capital, _i_t and _i*

t–1 are one-period interest rates, respectively

at home and abroad, _T_t denotes lump-sum taxes, and _Ξ_tstands for net payments from labour income

insurance. Function Φ(.) describes the risk premium associated with foreign transactions and is specified as follows:

(

)

Σ

(

)

( )

l wdj w l_t 1 _j_t, 0 1 t, j t t, j W W w w + 1 1 1

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

Σ

0_t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(6) where:

(

)

Σ

(

)

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

Σ

0_t₌₀ E

Σ

]

[

( ))

(

]

[

]

( )

(7)

is the ratio of net foreign assets to steady-state output _{y–, φ}_t, denotes a risk premium shock, while χ is

the risk premium elasticity with respect to foreign debt.

3.2. Firms

Producers of final goods. Final goods _y~_t, used ultimately in the home country for consumption,

investment and government purposes, are manufactured by producers of final goods in accordance with the following formula:

(

)

Σ

(

)

jt t j t j t t j t j _x x x S k k _, 1 , , , 1 , = 1 + 1 = = ≡ ≡ – – – – – – – – – – – – – +

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at 1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

(

)

(

)

∫

λ λ σ Ξ Φ Φ φ φ φφ ω η η η η η _ωη ηη φ ε ε –1 ε ε θ Λ Λ α α ε ε – 1 ε ε 0 t E

Σ

0 t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(8)

where _y_H,t_{and y}_F,tare Dixit-Stiglitz aggregators of local and foreign intermediate goods:

(

)

Σ

(

)

jt t j t j t t j t j _x x x S k k _, 1 , , , 1 , = 1 + 1 = = ≡ ≡ – – – – – – – – – – – – – +

( )

/

_t t t P P

(

t

)

t t

(

t

)

(

t t

)

(

at 1, t

)

exp

[

(

at1 t

)

]

(

)

+ = 1 1 1 1 1 1 _H_,_t _F_,_t t y y y~

{

H,F

}

X , di y yXt,= Xi,t, = = 1 1 0 1 1 0 di y y * t, i, H * t, H 1 t, i t, i t t, i, H zk l y

(

–

)

(

+

)

P

h h , , 1 , 1 , , + = 1 , , ,t= Ht Ht H

P

(

F,,it t t*

)

(

)

[

(

)

]

{

}

( )

0

,

1 N

~

ˆ

x,t t , g , ,lt, t, t, z , t g , t mt,, y , *t t* t y , y~ , t c , t x , t l , t k , t r , t w , t p , kt, t, Ht,, F,t, t, t, mc , t s , t q , t at

(

)

(

)

∫

λ λ σ Ξ Φ Φ φ φ φφ ω η η η η η _ωη ηη φ ε ε –1 ε ε θ Λ Λ α α ε ε – 1 ε ε 0 t E

Σ

0 t = 0 E

Σ

]

[

( ))

(

]

[

]

( )

(9)

while _{ω is the share of foreign goods in the basket of goods consumed in the home country.}1

The elasticity of substitution between domestic and foreign goods is defined by parameter _{η, while}

the elasticity of substitution between intermediate goods is _ε.

1_{The model does not take account of the option of re-exporting imported goods. We consider this fact while calibrating} the ω parameter.